SpliceJumper

SpliceJumper applies a classification-based machine learning approach to identify splicing junctions from RNA-seq data for accurate transcriptome profiling and alternative splicing analysis.

Key Features:

• Classification-Based Approach: Employs machine learning models to distinguish true splicing events from sequencing artifacts using extracted features from RNA-seq data.
• Feature Extraction: Extracts multiple data-derived features from RNA-seq reads for use as model input.
• Model Training and Classification: Trains classification models on extracted features to classify potential splicing junctions.
• Comparative Performance: Demonstrated superior accuracy relative to TopHat2 and MapSplice2 on both simulated and real datasets.
• Improved Splice-Site Mapping: Enhances accurate mapping at intron–exon junctions to reduce incomplete read mapping.

Scientific Applications:

• Transcriptome Profiling: Enables more accurate genome-wide identification of splice junctions for transcriptome characterization.
• Gene Structure and Expression Studies: Supports analysis of gene models and transcript abundance by improving junction calls.
• Alternative Splicing and Isoform Analysis: Facilitates detection and characterization of alternative splicing events and isoform diversity.
• Functional Genomics and Disease Research: Aids construction of transcript variant models relevant to functional genomics and diseases associated with splicing anomalies.

Methodology:

Extracts multiple features from RNA-seq data and trains machine-learning classification models to distinguish true splicing junctions from sequencing artifacts; performance was evaluated on simulated and real datasets against TopHat2 and MapSplice2.